Cloud workloads combine software components from different parties to process sensitive data. Each component has its own trust model - it must protect its assets from the rest of the system, yet share sensitive data with components it cannot trust to keep confidential. This tension requires composing isolation boundaries for confidentiality and encapsulation. Unfortunately, the cloud offers no direct way to compose such boundaries, forcing tenants to assemble, deploy, and maintain their own solutions. This paper shifts that burden back to the infrastructure by making composable, attestable isolation a first-class systems abstraction. We present Tyche, a security monitor that centers isolation around a unified composable abstraction: security domains (SDs). An SD is an execution environment whose access to machine resources - memory, cores, devices - is controlled through explicit capabilities. A small set of capability operations enables SDs to partition, share, and reclaim resources; by nesting recursively, SDs compose attestable trust boundaries for confidentiality and encapsulation. Tyche attests these compositions, providing end-to-end security guarantees for workloads made of mutually distrustful components. As a first-class cloud primitive, this single abstraction subsumes enclaves, sandboxes, CVMs, and their compositions. Tyche provides composable isolation without sacrificing compatibility with existing hardware and software stacks. It runs on commodity x86 64 hardware without security extensions, and a RISC-V prototype demonstrates portability across platforms. Our SDK composes isolation for unmodified workloads within SDs with minimal overhead. In a confidential LLM inference scenario with mutually distrustful users, model owners, and cloud providers, the slowdown is just 2% compared to bare-metal Linux.

In this article, we investigate the multiagent robust consensus problem under model uncertainties, where the uncertain matrices and initial values are bounded by prior intervals. Based on the positive system theory, the related upper and lower dynamic systems are constructed to guarantee that the state value remains within a specified range. Subsequently, in accordance with the Lyapunov stability principle, the observation and consensus errors converge to zero, that is, the real states are reconstructed and consensus is achieved. Both local and neighborhood protocols, which are utilized to realize robust consensus, are presented. Notably, the proposed methods increase the design freedom and eliminate the Metzler constraint on the error matrix by introducing two novel parametric matrices. Without loss of generality, the topology in this article is assumed to contain a directed spanning tree, which can be directly degenerated to the undirected graph. Finally, numerical simulations validating the theoretical results are described.

Software is going through a trust crisis. Privileged code is no longer trusted and processes insufficiently protect user code from unverified libraries. While usually treated separately, confidential computing and program compartmentalization are both symptoms of the same problem, deeply rooted in hierarchical commodity systems: privileged software's monopoly over isolation.

In this paper, we propose a text-guided image manipulation method which focuses on editing shape attribute using text description. We combine an image generation model, StyleGAN2, and image-text matching model, CLIP, and we have achieved the goal of image shape attribute manipulation by modifying the parameters of the pretrained StyleGAN2 generator. Qualitative and quantitative evaluations are conducted to demonstrate the effectiveness of the proposed method.

Wireless cellular networks have many parameters that are normally tuned upon deployment and re-tuned as the network changes. Many operational parameters affect reference signal received power (RSRP), reference signal received quality (RSRQ), signal-to-interference-plus-noise-ratio (SINR), and, ultimately, throughput. In this paper, we develop and compare two approaches for maximizing coverage and minimizing interference by jointly optimizing the transmit power and downtilt (elevation tilt) settings across sectors. To evaluate different parameter configurations offline, we construct a realistic simulation model that captures geographic correlations. Using this model, we evaluate two optimization methods: deep deterministic policy gradient (DDPG), a reinforcement learning (RL) algorithm, and multi-objective Bayesian optimization (BO). Our simulations show that both approaches significantly outperform random search and converge to comparable Pareto frontiers, but that BO converges with two orders of magnitude fewer evaluations than DDPG. Our results suggest that data-driven techniques can effectively self-optimize coverage and capacity in cellular networks.

Molecule design is the process of discovering potential compound candidates for drug discovery. Deep learning technique shows significant advantages in data mining and can be used for molecule design. However, most drug discovery projects are limited to low-data situations, and it is difficult to train deep learning neural networks. This paper proposes a novel drug design system that is based on deep learning. It adopts one-shot learning and reinforcement learning, and it can operate under low-data conditions. Once trained, the system can generate new molecules with the desired properties.

A radio-frequency (RF) energy harvester collects the radiated energy from nearby wireless information transmitters. Multiple wireless transmitters concentrate their radiation on the RF energy harvester while satisfying the basic requirement of the information links. To achieve this, a deep learning method is proposed for the multiuser transmission. A deep neural network (DNN) is implemented in each wireless transmitter. The DNNs are trained offline with simulated channels and applied online to generate transmit covariance matrices that meet the communication requirement and approach the maximum sum received power at the RF energy harvester.

A multi-antenna wireless transmitter communicates with its information receiver while beaming the radiated power to multiple nearby radio-frequency energy harvesters. The transmitter knows the channel to the information receiver but not the ones to the energy harvesters. By designing its transmit covariance matrix, the transmitter maximizes the minimum harvested power among the multiple energy harvesters while maintaining the information rate toward the receiver. To achieve this, we introduce a simplified channel vector from the transmitter toward the energy harvester. It can be estimated through particular transmissions and very limited feedback from the energy harvester to the transmitter. Once the transmitter obtains the simplified channel vectors, it can find the optimal transmit power allocation. To avoid high computational complexity, we propose a method to find the optimal power allocation with a deep neural network instead of solving a convex optimization problem. The simplified channel vectors are the input to the deep neural network. The neural network is trained offline with a large number of simulated data. Simulation results validate the method and show its superior performance compared with the convex optimization approach.

Devices with the capability of radio-frequency energy harvesting can collect the radiated energy from adjacent wireless energy transmitters. If the multi-antenna transmitter knows the vector channel to the energy harvester, it can design an optimal transmit covariance matrix that satisfies the energy harvesting requirement. However, it is impractical for the energy harvester to estimate the channel. In this paper, we propose a method to design the wireless transmission with a neural network. The transmitter uses a set of special beam patterns and the energy harvester measures the received power and feeds the power values back to the transmitters. The neural network then takes in the power values and outputs the transmit covariance matrix that can meet the energy harvesting requirement. The neural network is trained offline with a large number of simulated data. Simulation results validate the proposed method and show better performance than other wireless energy transmission methods.